SOCIO-DEMOGRAPHIC PROJECTIONS TO 2020 FOR ANALYSIS OF CHILD POVERTY IN UK REGIONS

Transcription

1 SOCIO-DEMOGRAPHIC PROJECTIONS TO 2020 FOR ANALYSIS OF CHILD POVERTY IN UK REGIONS John Parsons and Phil Rees School of Geography, University of Leeds, Leeds LS2 9JT Paper to be presented in the Child Poverty Session organised by Helen Barnard, Joseph Rowntree Foundation, York at the Work, Pensions and Labour Economics (WPEG) Conference 2006, University of Kent, Canterbury, July ABSTRACT The Joseph Rowntree Foundation has commissioned a set of studies that evaluate the feasibility of the Government s ambitious targets for reducing child poverty ( This paper reports on one of those JRF studies. The aim of our study was to supply projections of key variables related to child poverty for 2010 and 2020 for the UK and its regions. These are being used to re-weight analyses by the Institute of Fiscal Studies (IFS) and to enable some assessment to be made of the contribution of changes in size of the child population and its composition in terms of risk related categories. The IFS researchers are employing recent Family Resources Survey data in a micro-simulation model that tracks family poverty. In the paper we describe how we projected a set of tables of the future population distribution by seven child poverty linked variables across 13 regions. The variables are age and dependency status (to identify the numbers of dependent children), the number of dependent children in families (the higher the number the greater the risk), type of family (lone parent families face higher risks), size of household (larger households are at greater risk), ethnic group (there is considerable variation in child poverty rates by ethnicity), number of earners (directly related to family poverty) and housing tenure (social and private renters are poorer). Acting pragmatically, we have used three different approaches to projection: (1) adopting official projections (for age, some household categories, employment forecasts), (2) extrapolating inter-census trends or FRS statistics (number of dependent children, type of family, size of households, number of earners, housing tenure) and (3) developing our own cohort-component projections (ethnic group). For each region we projected seven sets of marginal tables, each containing number of people and each summing to the most accurate total, the total regional population estimated by combining GAD national (2004-based) and ONS sub-national (2003- based) projections. This ensured consistency of the key variable tables. We have also used these seven regional variable population vectors to update a multidimensional array based on the 2001 Census Individual Sample of Anonymised Records, though the results are not fully satisfactory because we are using data from different sources. In the paper we describe the demographic, socio-economic and ethnic profiles of the population in 2010 and 2020 across 13 UK regions and discuss their implications for child poverty reduction. i

2 1. INTRODUCTION Poverty is defined by Department of Work and Pensions (DWP) and HM Treasury (HMT) as living on less than 60% of the UK average (median) income. The median version of the definition is usually used. Child poverty is the number of children living in households with net household income below 60% of the median. The household income needs to be adjusted for the size and composition of the household. Larger households need more income than smaller to be above the poverty threshold. A clear exposition of how this equivalence of household incomes is achieved at present and the forthcoming changes in the methodology is described in Brewer et al. (2006), Appendix A. In 1999 the Prime Minister, Tony Blair pledged government action to eradicate child poverty by Brewer et al. (2006) interpret the DWP and HMT targets to be a 50% reduction from 1998/9 child poverty numbers by 2010, following a target of 25% reduction in numbers by 2004/5. No precise meaning has been attached to eradicate as yet; perhaps a further 50% reduction in numbers could be anticipated from 2010 to The percentage of children in relative poverty in 2004/5 was estimated to be 27.2% (Before Housing Costs) and 19.5% (After Housing Costs) (Brewer et al. 2006, Table 3.3). A set of forecast populations for 2010 and 2020, to be used as constraint tables, were requested by the Joseph Rowntree Foundation (JRF) as input to their project on UK child poverty. The forecast populations have been used by the Institute for Fiscal Studies, London (Mike Brewer and colleagues) to re-weight, to 2010 and 2020, the results of a tax and benefits micro-simulation model that explores policy options for eradicating child poverty. In order to estimate the projected trajectories we used 2001 Census regions. Income levels and poverty vary considerably between UK regions. In the 1999 to 2002 Expenditure and Food Survey regional average incomes ranged from a low of 22% below the UK mean to 33% above (National Statistics 2004, Table 8.2). Using regions as a means of disaggregating the population had computational advantages. We could apply our iterative proportional fitting model to thirteen smaller populations of about 1

3 4.5 to 5 millions rather than one larger UK population of 59 to 64 millions. It also means we could examine regional differences in the risk factors affecting child poverty. Each of the seven child poverty-linked variables will be considered in turn, describing the projection method, charting the trend at UK level and then examining differences at regional level using a special cartogram designed for demographic mapping (Dorling et al., 2003). An 8-dimensional array (the 7 variables plus region) was constructed by computationally combining the constraint tables. The initial 2001 population was based on the Individual Sample of Anonymised Records (ISAR) derived from the 2001 Census of Population (ONS, 2005). That process is outside the scope of this paper but is described in our Final Report to JRF (Rees and Parsons, 2006). Data sets generated by many different government departments were employed in the projection work (see Acknowledgements for full citations). Population and household counts from the past three censuses were used along with population estimates for census years. Census data are produced by the Office for National Statistics (ONS), General Register Office for Scotland (GROS) and the Northern Ireland Statistics and Research Agency (NISRA) (see ESRC/University of Manchester 2006, National Statistics/University of Durham 2006, GROS 2006 and NISRA 2006). Population projection information for the countries of the UK was sourced from the Government Actuary s Department s national projections (GAD 2005b) and the ONS sub-national projections for England (ONS 2004). Household data and the interim 2002-based projections were accessed from the Office of the Deputy Prime Minister for England (ODPM 2004). The most recent 2004-based household projections (ODPM 2006) were published too late to be incorporated in our work. The main difference in the 2004 projections, apart from the higher numbers consequent on revised national and sub-national population projections was an even higher projection of one person households (mainly aged), which does not directly affect child poverty. Household projections for Wales were sourced from National Assembly for Wales (NAW 2004) and for Northern Ireland 2

4 from NISRA (2004). The long-term labour force participation projections of HM Treasury (2004) were used to inform our projections of number of earners in households. The Family Resources Survey (FRS) is used to provide trends in family and household composition (DWP 2006 and ESDS Government 2006). 2. AGE AND DEPENDENCY STATUS The age and dependency status tables are the most important of the 7 constraints as all the other tables were adjusted to add up to these population totals. Data on projected populations by country 1 (from GAD 2005b) and by Government Office Region (GOR) within England (from ONS 2004) were assembled in spreadsheets for processing. The national population projections use a single-region cohort-component model with single years of age for four countries connected by net migration flows with net external migration assumptions. The sub-national population projections for England use a multi-region cohort component model for local authorities. We selected age groups that were a compromise between those available in the FRS and those available in the ISAR. The age groups were 0-9, 10-15, dependents, nondependents, 20-24, 25-29, 30-44, and Projection method The relevant parts of these data were loaded into a spreadsheet and the following operations performed. (1) Populations were aggregated to the target age groups, using a mixture of 5 year and single year ages with an assumption about the 15, split for the England GORs and London Boroughs. The England proportions these two groups making up the age group were employed. Figures were directly available for ages 15 and for Wales, Scotland and Northern Ireland from GAD (2005b). (2) The London Borough population projections were summed to yield Inner London and Outer London totals. (3) The age group was split between dependent and non-dependent children using a labour force participation rate computed from the 2001 Census and an assumption that these rates would increase only marginally by The latest Department for Education and Science statistics suggest that participation in Full Time Education among year olds is currently on a flat plateau, 3

5 despite government efforts to raise the rate. At most, we predict a 1% rise over the next 5 years, and 3% by 2020 (i.e. 0.2% per annum). (4) The ONS GOR and Inner and Outer London projections were adjusted to agree with the England projections of GAD. (5) The Country and GOR populations were adjusted by small amounts to add exactly to the UK populations in the 2001 estimates and in the 2010 and 2020 projections. 2.2 Resultant UK trends The UK population, shown in Figure 1, is still growing in the period from 58.8 millions in 2001 to 61.6 millions in 2010 and 64.4 millions in The numbers aged 0-15 decline from 11.9 millions in 2001 to 11.3 millions in 2010 but recover a little to 11.4 millions in However, the reduction compared with 2001 should make it slightly easier to achieve child poverty reduction goals. Note that the and 60+ age groups increase in population between 2001 and 2020, reflecting the ageing of the population. Combining age groups 0-9, and 16-19d shows that the total number of dependent children will decline slightly, while other age groups steadily increase (Figure 2). The ratio of dependent children to non-dependents will decline from in 2001 to in Resultant regional trends The national population growth is unevenly distributed across countries and regions as Figure 3 demonstrates. Scotland (SC) and the North East (NE) grow very little between 2001 and The North East faces minor decline after The North West (NW), Yorkshire and the Humber (YH), West Midlands (WM), Wales (WA) and Northern Ireland (NI) will experience lower than average UK growth to both 2010 and The East Midlands (EM), East (EA), South East (SE), South West (SW) and Outer London (OL) will grow faster than the UK. The highest growth is expected in Inner London (IL), though this will depend on the provision of sufficient new housing to house the extra people. 4

6 2.4 The impact of regional trends on child poverty One of the questions being addressed by the Child Poverty investigation of the Joseph Rowntree Foundation is whether trends in the population distribution and composition will help or hinder the attainment of the goals of child poverty reduction proposed by the UK government. Brewer et al. (2006, Appendix D) set out a methodology for carrying out such an analysis based on earlier work by Sutherland et al. (2003, Appendix II). Three effects are distinguished: the population effect, which depends on the change in the number of child dependents and the mean poverty risk, the incidence effect, which depends on how the poverty risk changes for particular family types and the compositional effect, which depends on how the mix of family types changes either favouring or disfavouring poverty reduction. The tables of projected persons by child poverty risk factors described in this paper will contribute to an analysis of the population effect and the compositional effect. Some comments on the likely impacts of the seven poverty risk variables can be offered here. The full analysis will be published in summer 2006 by the Joseph Rowntree Foundation. Table 1 gives some indication of the overall population effect. The numbers of children will decrease by 0.2 million between 2004/5 and 2010/11, so this will contribute only a little to the required decrease of 1.3 million in children in poverty. Between 2010/11 and 2020/21 the population of children increases marginally by 0.04 million (due to a subdued second baby boom echo children of the children of the baby boomers). The final column of Table 1 also shows how challenging are the government s targets set out in bottom part of the middle column. In 1998/99 when the policy goal was announced, a shocking 32.5% of children lived in poverty; by 2004/05 this had been reduced to 27.2%, just missing the government s goal of a 25% reduction in numbers (Brewer et al. 2006). The target for 2010/11 is for the number of children in poverty to be 50% of the 1998/99 figure, which translates, when projected numbers of all children are fed in, a poverty rate of 17.2% (10% lower than 2004/5). Will the shifts between regions of the population help contribute to this challenging target? Table 2 sets out the numbers of children by regions ranked by median weekly income after housing costs shown in the second column. We use the median after housing costs statistic to obtain a better picture of income in London where a 5

7 concentration of very high earners pulls up the mean income and where housing costs are high. The third to fifth columns in Table 2 report the numbers of children in each region. These are expressed as a percentage of the Great Britain total in columns six to eight. The final two columns show the shifts that are projected to occur in this decade and the next in the regional distribution of children. Overall the children share of regions with above GB average income will increase (South West, Inner London, Outer London, East and South East) and regions with below GB average income will see a decrease (North East, Yorkshire and the Humber, Wales, West Midlands, the North West and Northern Ireland). There are two exceptions. The East Midlands has below GB average income levels but sees an increase in its share of the GB child population. The East Midlands has two sub-regions: Northamptonshire, Leicestershire and southern Lincolnshire are really the outer parts of the London region and benefit from the jobs and migrants that its economy exports; Nottinghamshire, Derbyshire and north Lincolnshire are part of northern industrial England, facing the same challenges of renewal after industrial restructuring as its northern cousins. The major anomaly is Scotland, with higher than GB average income but the high losses of children. This position reflects the drop in fertility in Scotland over the past two decades to a position near the UK level to well below that level, coupled with favoured treatment by the national fiscal system which supports a higher level of public expenditure than elsewhere in the UK. Overall, we would expect a modest contribution of regional population shift to the poverty reduction goal. 3. HOUSEHOLD SIZE The household size constraint table had to be created next as it was used to help convert many other variables from household counts into population counts. In its original form, as a count of households, the values were 1, 2, 3, 4, 5 and 6+ persons. When converted to a population array (count of people in households of that size), the additional categories Not in Household and Visitors had to be added to ensure consistency with the ISAR. 3.1 Projection method The following steps were used to project the household size distributions for countries and regions. 6

8 (1) Relevant data were assembled from past Censuses and ODPM/GROS/NAW projections 3. For NI, official projections for all household sizes for both years are provided by NISRA, so steps (2) to (5) were not needed for this region. (2) For other regions, only one-person household figures are available (usually for 2011 and 2021), so these have to be scaled to 2010 and 2020 first. (3) From the 2001 Census, we obtained regional household totals for household sizes 1 to 6+. (4) The base totals were projected linearly using the Census trend, as this seemed to best match the assumptions behind ODPM s one-person household projections. (5) The one-person household figures from ODPM/GROS/NAW were substituted for the estimates and the remaining proportions rescaled to equal 100% for household sizes 1 to 6+ inclusive. (6) The household matrix was converted into a people matrix. (7) The projections are adjusted take account of the Not in Household and Visitor category proportions in the 2001 ISAR (which are held constant). (8) The overall population totals are constrained to the Age and Dependency Status constraint table total. 3.2 Resultant UK trends Figure 4 shows that there will be a continued strong increase in 1-person and 2-person households, a slight increase in 3-person and a decline in 4-, 5- and 6+-person households. This is most likely result of both people choosing to have smaller families and more people living on their own out of choice, because of relationship break-up or through loss of spouse or partner in old age. Other things being equal, smaller families should make the targets for reducing child poverty easier to attain. 3.3 Resultant regional trends Figure 5 shows a conventional UK regional map and a cartogram equivalent we have used the latter throughout, as the area of the region on the page is proportional (roughly) to its total population. It makes each region roughly the same size and easier to compare visually. We downloaded the Great Britain cartogram (Dorling et al., 2003) and added Northern Ireland for full UK representation. The maps show 12 of the 13 regions used with Inner and Outer London not distinguished. Note that full 7

9 results for 13 regions are provided in Rees and Parsons (2006). Figure 6 selected the largest household size category, 4+ persons, with the highest risk of containing children in poverty. We project (Figure 6) that Northern Ireland (-12.7%), London (-11.2%), West Midlands (-8.9%) and East of England (-7.8%) will see the biggest drop in large (4+ person) households. Northern Ireland will still have more large households than any other region, but it will gradually converge towards the norm. 4. DEPENDENT CHILDREN A key constraint is the number of people in households with dependent children. The values for this variable are Not in household and households with 0, 1, 2 and 3+ dependent children. The 3+ upper bound was imposed by Census/ISAR data as the FRS reports households with up to 6+ children, though these are rare. 4.1 Projection method The steps used to produce these projections were as follows. (1) We assembled data files from past Censuses (1981, 1991 and 2001) and past/current FRS data (1998/99 to 2004/05). (2) From the 2001 Census, we extracted regional household totals for each category of dependent children in household. (3) The household totals were increased linearly from the 2001 Census base using the FRS trend, as this seems to be good match for both the longterm and short-term trends. (4) The projected totals are then used as marginal column totals in an IPF routine which adjusts a matrix generated from the ISAR with the marginal row totals taken from the Household Size constraint table. The result is a household matrix, consistent across these 2 dimensions. (5) The household matrix is converted into a people matrix. (6) The population totals are adjusted to include the non-household population in the 0 dependents total. (7) The overall population totals are constrained to The Age constraint table. 4.2 Resultant UK trends We project (Figure 7) that the number of people in 1- and 2-child dependent households will stay much the same; that is, the number of persons 0 child dependent 8

10 households will increase slightly and the number of persons in 3+ child dependent households will decline slightly. This trend is most probably associated with decreased fertility and therefore smaller family sizes. 4.3 Resultant regional trends Figure 8 shows the regional distribution of the percentage of people in households with 3 or more dependent children. All regions show a slight decline, especially Northern Ireland (-2.6%), London (-2.1%), West Midlands (-2.1%) and North West (- 1.7%). 5. FAMILY TYPE The Family Type categories do not match Census or FRS categories exactly, but have been chosen because official ODPM regional projections already exist for them. The values are: Not in family/household, One-person, Lone Parent, Married Couple (with or without children), Cohabiting Couple (with or without children) and Other Multiperson (e.g. student households). 5.1 Projection method The steps used to project persons by family type were as follows. (1) We extracted relevant data from the 2002-based interim projections for England (ODPM, NAW and GROS. For NI, official projections for household size only are provided, so (other than for one-person families), the Great Britain trend has to calculated and applied, before going to step 6. (2) For other regions, a good set of family type figures are available. They are usually for 2011 and 2021, so must be scaled to 2010 and 2020 first. (3) The projected totals are then used as column marginals in an IPF routine (with the row marginals taken from the Household Size constraint table). The result is a household matrix, consistent across these 2 dimensions. (4) The household matrix is converted into a people matrix. (5) The projections are adjusted take account of the proportion in the category not in a family in the 2001 ISAR, which is held constant. (6) The overall population totals are constrained to the Age and Dependency Status constraint table total. 9

11 5.2 Resultant UK trends ODPM predict (Figure 9) a proportional increase in One-person, Other multi-person and Cohabiting Couple household populations. The 7.8% decline in Married Couples does not come as a surprise but the 3.8% decline in the Lone Parent populations does. 5.3 Resultant regional trends Figure 10 shows the percentage of people living in lone parent households in 2001 and All regions, other than London (+2.0%) show a decline in Lone Parent households. This trend should be helpful in achieving reductions in child poverty. 6. ETHNIC GROUP Between 1991 and 2001, half of the UK s population growth was due to immigration. Population falls in some areas have only been prevented by increased migration from abroad. The ethnic group populations were calculated from scratch as planned ONS projections will not be available until mid-2006, although the methodology has been tested out in producing experimental ethnic group estimates for 2001 to 2003 (Large and Ghosh 2006). Below is a précis of the full method described in Rees (2006). The selected ethnic groups were a compromise between those available in the FRS, those available in the ISAR plus those used in other studies (Coleman and Scherbov, 2005) to better evaluate the results. The ethnic groups are: White, Mixed, Asian, Black and Chinese & Other. There is a case for breaking down the Asian category into Indian, Pakistani, Bangladeshi and Other Asian, as the Indian population is generally much wealthier than the Bangladeshi, but this additional disaggregation is not attempted here. Table 3 shows how fast ethnic minority populations having been growing between 1981 and Without this growth the population of England would have stagnated rather than growing by 5% over the two decades. The White British population (the major component of the White group but the least demographically dynamic) has probably decreased by a small amount over the twenty years. Ethnic minorities grew by 96% in the same period. 10

12 6.1 Projection method To avoid problems with small cell counts, a single region model was adopted for the current task. The model is a component model using period-cohorts. Period-cohorts are age-time spaces in which demographic components add to or subtract from the starting population. For example, the 20 year old population at the start of a projection interval becomes the 21 year old population one year later, having lost members through mortality, internal out-migration and emigration and having gained members through internal in-migration and immigration. For the current model all these variables must be estimated for each ethnic group. The notation used is as follows: P = population, P F = final population in a periodcohort, P S = start population in a period-cohort, s = survival probability, o = internal out-migration probability, e = emigration probability, M = internal in-migrants and I = immigrants. The indexes/subscripts (in parentheses) are defined as c = period-cohort (age x at time t to age x+1 at time t+1), g = gender (sex), t = time at start of interval, t+1 = time at end of interval, one year later, for flow components t means the interval starting with time t, i.e. to to t+1. The cohort-component projection is an implementation of the following model for each region. For ages 0 to 100+, the population of each ethnic group and gender is projected as: P P s P o P e P M F S S S S + + ct ct ct ct ct ct ct ct ct = (1) where P F ct is the final (end of interval) population of period-cohort c at time t, P S ct is I ct the start (of interval) population period-cohort c at time t, sct is the probability of survival for persons in period-cohort c in the time interval starting with t, oct is the probability of out-migration (within the country) for persons in period-cohort c in the time interval starting with t, ect is the probability of emigration (out of the country) for persons in period-cohort c in the time interval starting with t, M ct is the number of in-migrants from elsewhere in the country who are in period-cohort c in the time 11

13 interval starting with t, I ct is the number of immigrants from outside of the country who are in period-cohort c in the time interval starting with t. Prior to the start of the next time interval, final populations are transferred to become start populations for ages 0 to 99: P P S F = (2). c + 1 t+ 1 ct For the open ended age 100+, the final populations of the last two period-cohorts are added together P P P S F F = t t 99t + (3). Births to each ethnic group are projected by multiplying estimated age-specific fertility rates by an average population at risk of women in the fertile ages (15 to 46 in this case): B a = 46 = 0.5 { P S + t a 15 at = P F at } (4). A sex proportion, w g is applied (male proportion of 0.516, female proportion of 0.484) to the births to yield male and female births to generate the start population for the first, infant period-cohort from birth to age 0 P w B S = g (5). 1 t t The survival and migration equation is then used for the infant period-cohort: P P s P o P e P M F S S S S + + 1t 1t 1t 1t 1t 1t 1t 1t 1t 1t = (6) I The following steps were executed to estimate input variables for the projections. (1) The base populations of ethnic group populations by age were estimated from 2001 Census standard and commissioned tables. International migration flows (immigration and emigration) were estimated from the revised Total International Migration series of ONS. (2) Fertility rates for regions and ethnic groups were estimated from regional age specific fertility rates for the whole population and from 2001 Census child/woman ratios specific to each ethnic group. A similar technique was used by Large and Ghosh (2006). 12

14 (3) Births were projected using estimated fertility rates for each ethnic group and a sex proportion was applied to derive male and female births. Fertility rates were held constant over the projection interval 2001 to (4) Deaths are modelled as the start population multiplied by a period-cohort survival probability. Survival probabilities were assumed to increase as a result of a mortality decline at each age of 2% per annum (as observed on average over the last three decades). (5) The survival probability is derived from a UK life table adjusted to regional mortality conditions by applying a standardised mortality ratio. The survival probability for a region and gender = [1 {SMR/100} {1 survival probability for the UK and gender}. This technique shifts the mortality by age schedule up or down uniformly. Improved estimates could be developed employing regional deaths by age. The same survival probabilities were applied to each ethnic group. There is evidence (e.g. from the Longitudinal Study linking a sample of individuals through four censuses) that there are significant ethnic mortality differences but a definitive study identifying regional and ethnic differences has not yet been done. (6) Out-migrants are modelled as the start population multiplied by a periodcohort survival probability (see equation 1). Estimates were made of ethnic internal and external migration from 2001 Census Standard and Commissioned tables and Total International Migration tables (see Rees 2005 for details). (7) Emigrants are modelled as the start population multiplied by a period-cohort survival probability (see equation 1). (8) Internal in-migrants and immigrants are entered as estimated flows into projection equation (1). (9) The ethnic population totals are constrained to the GAD projection totals for each country. (10) The overall population totals are constrained to the Age and Dependency Status constraint table total. 6.2 Resultant UK trends Figure 11 plots the dependent and non-dependent populations for all ethnic groups. The group White dependents is the only category to decline over time because of 13

15 declining births of Whites resulting from continuing low fertility rates and smaller numbers of women in the fertile ages. Black and Chinese and Other dependents don t increase much; whereas Mixed and Asian dependents show a strong increase. The biggest increase is in Asian non-dependents, from 1.5 to 2.5 millions. This merits further examination. Figure 12 shows the biggest growth derives from the 60+, and age groups, respectively much stronger than the UK as a whole (Figure 1). 6.3 Resultant regional trends Figures 13 and 14 show the White and Asian composition of UK regions in 2001 and 2020 (these are the two biggest ethnic groups). The White percentages decrease in all regions and the Asian percentages continue to increase. London and the south east in general continue to see the greatest change due to the region s capacity to create jobs, which attracts international migrants to both established communities and helps fashion new migrant communities. International immigrants have largely concentrated in the UK s biggest cities and particularly in southern metropolises in the past two decades whereas in the 1950s to 1970s they migrated more strongly to Midland and Northern cities. They are also experiencing redistribution through internal migration, with some signs of net shifts to suburban and metropolitan rings in the London region and of shifts from the less vigorous economies of northern cities to southern (Rees and Butt 2004). 7. NUMBER OF EARNERS We also examined and projected trends for the population in households with a given number of earners. This is probably the most speculative projection. The categories for the variable are Not in household, 0, 1, 2 and 3+ earners. 7.1 Projection method The steps in projecting this variable distribution were as follows. (1) We assembled data files from 1981, 1991 and 2001 Censuses tables and from the 2001 Census ISAR and used the HM Treasury projections for total number of employed adults. (2) We constructed a base population from 2001 Census. 14

16 (3) The household matrix was converted into a people matrix. (4) The projections are increased linearly from the 2001 Census base, in line with the overall Treasury trend, maintaining the number of employed adults/total people ratio. (5) The overall population totals are constrained the Age constraint table. 7.2 Resultant UK trends Figure 15 shows that, in 2010, there is a slight proportional increase in earners, but in 2020 it slips to back under the 2001 level (proportionally). 7.3 Resultant regional trends Figure 16 shows there is expected to be very little change, proportionally, between 2001 and 2020 (about +0.4% in all regions). 8. HOUSING TENURE Values for tenure are Not in household, Private rent, Social rent and Owns. Social housing provision varies greatly by region and corresponds well to the numbers in poverty in that area. However, the decline in social housing does not necessarily reflect increased wealth. Shortages of social housing have led to long council waiting lists and overcrowding in some towns, contributing towards poverty. Government strategy is now to force house-builders to include minimum numbers (30-40% depending on local authority) of affordable/social housing on new developments over 14 units in size. 8.1 Projection method The steps used to project the tenure distribution were as follows. (1) Data from 1991 and 2001 Censuses and the 2001 Census ISAR were assembled together with NISRA projections. (2) A base population was constructed from the 2001 Census. (3) The household tenure proportions by region were changed linearly from the 2001 Census base using the NISRA trend for Northern Ireland and the Census trend for all other regions. (4) The household matrix was converted into a people matrix. 15

17 (5) The overall population totals were constrained to the Age and Dependent constraint table total. 8.2 Resultant UK trends ODPM (2004) projects (Figure 17) that the biggest change in tenure will be the continued decline of social housing (-4.8%), following by the rise of private renting (+4.2%) and a slight increase in owner occupation (+0.5%). 8.3 Resultant regional trends Figure 18 shows London and the North East continue to have the most social housing, with Northern Ireland and Yorkshire & Humberside third and fourth, respectively by However, London is the only region where ODPM predicts that social housing will increase (+5.3%). Elsewhere, 5-10% drops are the norm. 9. CONCLUSIONS The government has failed to achieve its target of reducing child poverty by 25% between 1998/99 and 2004/05 (DWP, 2005). It fell from 4.1 million to 3.4 million (17%). The IFS press release accompanying publication of Poverty and Inequality in Britain: 2006 (Brewer et al. 2006) points out that, child poverty will have to fall half as quickly again over the next six years as it has done over the past six years (IFS 2006, p2). This will be a tough target, even though our calculations show the child population declining from 13.1 millions in 2004/5 to 12.8millions in 2010 and only increasing slightly between 2010 and 2020 (see Table 1). Under the government s welfare-to-work policies, more than 300,000 extra lone mothers have found employment, but campaigners believe these strategies have left behind large families or those with disabled children. Philip Hammond, Shadow Work and Pensions Secretary said We must create more opportunities for parents who are struggling to get back into work, (BBC, 2006). HM Treasury forecasts that the number of people in employment will rise in 2010 and 2020 but not as much as ONS forecasts the total population to rise. ODPM family 16

18 type forecasts are more optimistic in that they predict that the number of people in Lone Parent households will decline from 11% to 7.2% between 2001 and These and other socio-demographic trends will continue to affect the child poverty reductions that are achieved. Our constraint tables provide different slices of the UK population in 2010 and 2020 and are a useful set of inputs for the estimation of future likely child poverty levels by the Institute for Fiscal Studies. END NOTES 1 Country in this context means the UK constituent territories of England, Wales, Scotland and Northern Ireland, the governmental and legislative arrangements for which are partially devolved from the UK Parliament in Westminster. England is divided into nine Government Office Regions, including London. We have split London into two groups of boroughs Inner and Outer London. So in total the study uses 13 regions: North East, North West, Yorkshire and The Humber, East Midlands, West Midlands, East of England, South East, South West, Inner London, Outer London, Scotland, Wales and Northern Ireland. 2 The spelling dependent is used throughout for both the adjective and the noun, to avoid confusion. 3 At the time that the work reported here was carried out, the latest England household projections were the interim 2002-based projections (ODPM 2004). New 2004-based projections were published in March 2006 (ODPM 2006). The ODPM projections only provide information about one person households and all other sizes combined, so contain relatively little information directly relevant to child poverty investigation. 17

19 ACKNOWLEDGEMENTS The research report in this paper was supported by a grant from the Joseph Rowntree Foundation to the University of Leeds, managed by Helen Barnard. Mike Brewer of the Institute of Fiscal Studies provided helpful advice. The School of Geography, University of Leeds has an HMSO Click-Use Licence for re-use of Crown Copyright data for research and teaching purposes. Data Sources (Census) We provide here a summary of the citations required by the end user licences for each census dataset we used. Full details are given at Dataset: 2001 Census: Samples of Anonymised Records (Licensed) (England, Wales, Scotland and Northern Ireland) (ISAR) Census output is Crown copyright and is reproduced with the permission of the Controller of HMSO and the Queen's Printer for Scotland. Citation Original Data Depositors: ONS, GROS and NISRA, 2001 Census : Samples of Anonymised Records (Licensed) (England, Wales, Scotland and Northern Ireland) ISAR file provided by ESRC/JISC Census Programme, Cathie Marsh Centre for Census and Survey Research (University of Manchester). Datasets: 2001 Census: Standard Area Statistics (England and Wales) 2001 Census: Standard Area Statistics (Scotland) 2001 Census: Standard Area Statistics (Northern Ireland) 1991 Census: Local Base Statistics (England and Wales) 1991 Census: Local Base Statistics (Scotland) 1991 Census: Small Area Statistics (Northern Ireland) Census output is Crown copyright and is reproduced with the permission of the Controller of HMSO and the Queen's Printer for Scotland. Original Data Depositor: ONS Census: Standard Area Statistics (England and Wales), 1991 Local Base Statistics (England and Wales) Original Data Depositor, GROS Census: Standard Area Statistics (Scotland), 1991 Census Local Base Statistics (Scotland) Original Data Depositor, NISRA Census: Standard Area Statistics (Northern Ireland), 1991 Census: Small Area Statistics (Northern Ireland) Computer files supplied by ESRC/JISC Census Programme, Census Dissemination Unit, MIMAS (University of Manchester). Data Sources (non-census): All data are Crown copyright and are reproduced with the permission of the Controller of HMSO and the Queen's Printer for Scotland. Family Resources Survey: DWP (2006), ESDS Government (2006). Household Projections: ODPM (2004), NAW (2004), NISRA (2004). Population Projections, GAD (2005b), ONS (2004). 18

20 REFERENCES BBC (2006) Government misses poverty target. BBC News, 19 March URL: Accessed 10 April Bijak J., Kupiszewska D., Kupiszewski M., Saczuk K. (2005) Impact of international migration on population dynamics and labour force resources in Europe. CEFMR Working Paper 1/2005. Central European Forum for Migration Research, Warsaw. URL: Accessed 14 June Brewer, M., Goodman, A., Shaw, J. and Sibieta, L. (2006) Poverty and Inequality in Britain: Commentary No.101. Institute for Fiscal Studies, London. URL: Coleman, D. and Scherbov, S. (2005) Immigration and ethnic change in low-fertility countries towards a new demographic transition? Paper presented to Population Association of America Annual Meeting, Philadelphia, March 31-April , Session 98, 1 April. URL: accessed 14 June Dorling D., Durham H., Rees P., Nelson A. and See L. (2003) Universal Data Maps. Collection of Historical and Contemporary Census Data and Materials Unit 13. URL: Accessed 10 April DWP (2005) Households Below Average Income (HBAI) 1994/ /05. Department for Work and Pensions, London. URL: Accessed 10 April DWP (2006) Family Resources Survey. Department of Work and Pensions, London. URL: Accessed 10 April EDINA (2005) UKBORDERS Boundary datasets of the United Kingdom. Edinburgh University Data Library. URL: Accessed 23 October ESDS Government (2006) Family Resources Survey Datasets. ESRC/JISC Economic and Social Data Service, Universities of Manchester and Essex. URL: Accessed 10 April ESRC/University of Manchester (2006) CASWEB: Web Interface to Census Aggregate Outputs and Digital Boundary Data. ESRC-JISC/Census Dissemination Unit, Manchester Computing. URL (needs UK Higher Education/Further Education Athens account). Accessed 10 April GAD (2005a) Projected increase of 7.2m in UK population by News Release by Office for National Statistics and Government Actuary s Department, 20 October URL: based_national_population_projections.pdf. Accessed 23 October GAD (2005b) Projections database: 2004-based principal projections. URL: Accessed 10 April GROS (2006) Scotland s Census Results Online (SCROL). General Register Office for Scotland, Edniburgh. URL: Accessed 10 April Haskey, J. (2000) Projections of the population by ethnic group: a sufficiently interesting or a definitely necessary exercise to undertake? Population Trends, 102, URL: 19

21 Accessed 10 April Haskey, J. (ed.) (2002) Population Projections by Ethnic Group: A Feasibility Study. ONS Studies in Medical and Population Topics, SMPS No.67. The Stationery Office, London. URL: Accessed 10 April HM Treasury (2004) Long-term Public Finance Report: an Analysis of Fiscal Sustainability. Her Majesty s Treasury, London. URL: Accessed 10 April IFS (2006) Government misses child poverty targets. IFS Press Release, March URL: Accessed 10 April Kiernan, K. and Smith, K. (2003) Unmarried parenthood: new insights from the Millenium Cohort. Study. Population Trends 114, URL: Accessed 10 April Large P. and Ghosh K. (2006) A methodology for estimating the population by ethnic group for areas within England. Population Trends, 123, Spring 2006, URL: Accessed 10 April National Statistics (2004) Regional Trends. No Edition. Edited by P. Causer and D. Virdee, TSO, London. URL: Accessed 10 April National Statistics/University of Durham (2006) NOMIS: Official Labour Market Statistics. URL: Accessed 10 April NAW (2004) Statistics for Wales: Housing. National Assembly for Wales, Cardiff. URL: Accessed 14 August NISRA (2004) Household Projections for Northern Ireland: Northern Ireland Statistics Research Agency, Belfast. URL: ldexec.pdf. Accessed 14 August NISRA (2006) Northern Ireland Census Access (NICA). Northern Ireland Statistics Research Agency, Belfast. URL: Accessed 10 April ODPM (1999) Projections of Households in England 2021 (19 October 1999). use_ hcsp ODPM (2004) Interim Household Projections for English regions (2002-based). Office of the Deputy Prime Minister, London. URL: Accessed 14 Aug ODPM (2006) New Projections of Households for England and the Regions to ODPM Statistical Release 2006/0042. Office of the Deputy Prime Minister, London. Accessed 10 April ONS (2004) 2003-based Subnational Population Projections for England: Data and Methodology Issues. Office for National Statistics, London. Available as 20

22 2003snpissues.pdf. URL: Accessed 10 April Rees P. and Butt F. (2004) Ethnic change and diversity in England, Area, 36(2): Rees P. (2006) Estimating International Migration at Regional Scale for Ethnic Groups in the UK. Paper presented at the Workshop on International Migration organised by the Social Science Research Institute, University of Southampton and the Netherlands Interdisciplinary Demographic Institute, September 2005, University of Southampton. URL: Accessed 10 April Rees P. and Parsons J. (2006) Child Poverty in the UK: Socio-demographic Scenarios to 2020 for Children. Final Report to the Joseph Rowntree Foundation as part of their project on Child Poverty in the UK. School of Geography, University of Leeds, Leeds LS2 9JT. Simpson, S. (2004) Statistics of racial segregation: measures, evidence and policy. Urban Studies, 41(3), Sutherland, H., Sefton, T. and Piachaud, D. (2003), Poverty in Britain, York: Joseph Rowntree Foundation. URL: Accessed 9 April Wilson, T. and Rees, P. (2005) Recent developments in population projection methodology: a review. Population, Space and Place, 11,

23 Table 1: Child Poverty (GB) estimates and targets Year Mid-point of interval Children in poverty AHC Number of children (millions) Percent in poverty AHC (millions) 1998/99 FY /00 FY /01 FY MY /02 FY /03 FY /04 FY /05 FY MY /11 FY MY /21 FY Sources: 1. Brewer et al. (2006), Tables 3.3 and 3.4, p Rees and Parsons (2006), Appendix B1, p DWP/HMT target of 50% of 1998/99 number of children in poverty. 4. Assumed target of 50% of 2010/11 number of children in poverty. 5. Interpolated linearly between 2010 and 2020 MY projected populations. 6. Extrapolated linearly using 2010 and 2020 MY projected populations. Notes: Poverty = household incomes below 60% of median after housing costs (AHC). FY = financial year (1 April to 31 March, mid-point 30 September) MY = mid year (30 June/1 July). Table 2: Changes in the numbers and regional composition of children in relation to income, Great Britain 2001, 2010 and 2020 Income Numbers of Children % GB Change in % 2004/ per week 1000s 1000s 1000s % % % GOR/Country North East Yorks & Humb Wales West Midlands North West East Midlands South West Scotland Inner London Outer London East South East Great Britain Northern Ireland

24 Sources: Income = Median AHC income, Family Resources Survey, Table 2.6 in Brewer et al. (2006). AHC = After Housing Costs. Children = 2001 Census data and GAD and ONS projections, Rees and Parsons (2006). Table 3: Ethnic change in England Population (1000s) Change indexes Ethnic Group = = =100 White Black South Asian Chinese and Other All ethnic minorities All groups Source: Rees and Butt (2004), Table 3, p.178. Authors computations based on 1981 Census Small Area Statistics, 1991 Census Local Base Statistics, 2001 Census Key Statistics for Local Authorities, estimates from Rees and Phillips (1996), estimates from the Linking Censuses through Time project (Dorling, Martin and Mitchell, 2002 and All Census data are Crown Copyright. 23

25 Figure 1: Projected UK population by broad age, 2001, 2010 and 2020 Source: GAD (2005b) UK population (thousands) d 16-19n Age group Figure 2: Projected UK population by dependent children, 2001, 2010 and 2020 Source: GAD (2005b) UK population (thousands) Dependent children Non-dependents

26 Figure 3: Projected regional populations, 2010 and 2020 Source: GAD (2005b), ONS (2004) and authors computations 120 % of 2001 population NE NW YH EM WM EA SE SW IL OL SC WA NI UK Region/Country Figure 4: Projected UK population by household size, 2001, 2010 and UK population (thousands) Not in household No usual residents Household size 25

27 Figure 5: UK regions on a conventional map and on a population cartogram Source: EDINA (2005), Dorling et al (2003) SC Scotland, NI Northern Ireland, WA Wales, NE North East, NW North West, YH Yorkshire and the Humber, EM East Midlands, EA East of England, LO London, SE South East, SW South West Figure 6: Percentage of people in each region in 4+ person households, 2001 and